Opto-Electronic Engineering, Volume. 52, Issue 5, 250061(2025)

Hypergraph computed efficient transmission multi-scale feature small target detection algorithm

Yemei Sun1, Xueting Sang1, Yan Zhang1, Guorui Liu1、*, and Shuaiyu Chen1、*
Author Affiliations
  • 1[in Chinese]
  • 1School of Computer and Information Engineering, Tianjin Chengjian University, Tianjin 300380, China
  • show less
    Figures & Tables(12)
    Efficient transfer multi-scale feature small target detection algorithm based on hypergraph computing
    HOAF module
    Hyper computation module
    Distance computation part of the execution process
    Structure diagram of LDTG-Head and DTG module. (a) LDTG-Head; (b) DTG module
    Comparison between ordinary convolution and group convolution. (a) Ordinary convolution; (b) Group convolution
    Partial images of the VisDrone2019 dataset
    Comparison of target detection effectiveness in complex scenes on VisDrone dataset. (a), (b) Night environment; (c), (d) High-altitude view; (e), (f) Dense occlusion scenes
    • Table 1. Comparison results of AP and params of different algorithms on VisDrone dataset

      View table
      View in Article

      Table 1. Comparison results of AP and params of different algorithms on VisDrone dataset

      MethodAP/%mAP0.5/%Params/M
      PedestrianPeopleBicycleCarVanTruckTricycleAwning-TricycleBusMotor
      Faster R-CNN[33]20.914.87.351.029.719.514.08.830.521.221.8
      Cascade R-CNN22.214.87.654.631.521.614.88.634.921.423.2
      YOLOv539.031.311.273.535.429.520.511.143.137.033.27.0
      YOLOX[34]34.824.516.972.434.440.523.117.853.136.035.354.2
      YOLOv7[35]37.934.69.476.136.329.820.110.643.241.834.06.0
      YOLOv840.330.112.577.745.836.027.014.954.542.738.210.6
      YOLOv9[36]42.433.414.279.545.839.429.516.957.544.540.36.8
      YOLOv10[37]43.434.714.580.546.537.328.015.855.245.740.26.9
      Mamba-YOLO[38]41.031.711.579.344.233.226.113.856.043.038.05.7
      BDAD-YOLO[39]37.830.010.577.042.432.524.813.653.039.636.13.2
      HETMNet(ours)45.235.814.281.649.741.130.718.060.551.142.44.8
    • Table 2. Results of HETMNet's ablation experiment on the VisDrone dataset

      View table
      View in Article

      Table 2. Results of HETMNet's ablation experiment on the VisDrone dataset

      ModelETMFPNLDTG-HeadLAMPmAP0.5/%mAP0.5:0.95/%Params/MGFLOPsSize/MB
      Model 138.222.810.628.521.5
      Model 241.525.19.729.519.6
      Model 340.824.88.124.316.4
      Model 442.025.79.228.618.8
      Model 542.426.04.820.49.9
    • Table 3. ETMFPN ablation experiment results

      View table
      View in Article

      Table 3. ETMFPN ablation experiment results

      ModelmAP0.5/%mAP0.5:0.95/%Params/M
      +PAFPN38.222.810.6
      +GoldYOLO40.524.413.0
      +BiFPN40.924.97.0
      +ETMFPN41.525.19.7
    • Table 4. LAMP ablation experiment results

      View table
      View in Article

      Table 4. LAMP ablation experiment results

      Speed_upmAP0.5/%Params/M
      42.09.2
      1.2542.65.6
      1.3542.45.0
      1.4042.44.8
      1.4542.34.6
    Tools

    Get Citation

    Copy Citation Text

    Yemei Sun, Xueting Sang, Yan Zhang, Guorui Liu, Shuaiyu Chen. Hypergraph computed efficient transmission multi-scale feature small target detection algorithm[J]. Opto-Electronic Engineering, 2025, 52(5): 250061

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Article

    Received: Feb. 28, 2025

    Accepted: Apr. 8, 2025

    Published Online: Jul. 18, 2025

    The Author Email: Guorui Liu (刘国瑞), Shuaiyu Chen (陈帅宇)

    DOI:10.12086/oee.2025.250061

    Topics